|Publication number||US6845344 B1|
|Application number||US 09/642,268|
|Publication date||Jan 18, 2005|
|Filing date||Aug 18, 2000|
|Priority date||Aug 18, 2000|
|Publication number||09642268, 642268, US 6845344 B1, US 6845344B1, US-B1-6845344, US6845344 B1, US6845344B1|
|Inventors||Maureen A. Lally, Kenneth R. Goguen, Peter C. Lauterbach|
|Original Assignee||Emc Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (12), Referenced by (38), Classifications (17), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The invention relates generally to tools and systems for measuring the performance of mass storage systems, and more particularly, to methods and apparatus for developing, measuring, analyzing, and displaying the performance statistics of a plurality of disk drive elements controlled through a disk drive controller connected to a plurality of host computers.
As the size and complexity of computer systems increase, including the number of host computers and the number of disk drive elements, it becomes increasingly important to measure and understand the functions and parameters which affect the performance of the system. The performance of the system can be typically measured in terms of input/output (I/O) response times, that is, the time it takes for a read or write command to be acted upon, as far as the host computer is concerned, by the disk drive controller system.
It is well known, in the field, to measure, usually using a single parameter, the instantaneous or average response time of the system. Typically, a host computer outputs one or more I/O requests to the disk drive controller, and then measures the time for a response to be received from the disk drive controller. This time duration, while representative of the response of a specific read or write command to the disk drive system, is most often not representative of the actual performance which can be obtained from the system.
A similar distortion, not representative of system performance, can occur when average response time values are determined. For example, a disk controller, using a cache memory in its write process, can have substantially different write time responses depending upon the availability of cache memory. An average response (the average of, for example, a write where cache was available and one where cache was not available) would be misleading and meaningless.
The performance of a large storage system is particularly difficult to measure since more than one of the host computers, which connect to be disk drive controller(s), can operate at the same time, in a serial or in a parallel fashion. As a result, a plurality of disk drive elements, usually arranged in a disk drive array, operating in either an independent fashion, a RAID configuration, or a mirrored configuration, for example, can have a significant yet undetectable bandwidth or operational problem which cannot be addressed, or discovered, when commands are sent only from a single host computer.
In U.S. Pat. No. 5,953,689, issued Sep. 14, 1999, assigned to the assignee of this application, an improved method of time synchronizing a plurality of hosts operating in a variety of different configurations, and of issuing commands according to a prescribed sequence, was described. The method described in the above-identified patent features sending test requests, for the mass storage system, from a “main” host computer to each of a plurality of client host computers, executing at each host computer a test request sequence by sending commands to the mass storage system, accumulating at each host computer data regarding performance of the mass storage system, the data being in response to the requests or commands sent by each particular host computer, and sending, from each host computer to a master host computer, data regarding the performance of the mass storage system in response to the host generated commands. Significant data reduction techniques control and organize the data for later analysis.
While this system worked well, it had, at the time the application for the system was filed, specific limitations with regard to creating the test configurations and with regard to the flexibility of the analysis and processing once the statistics had been collected.
The invention relates to an improved method for measuring the system performance of a mass storage system having a plurality of disk drive storage elements controlled by a disk drive controller. Typically the disk drive controller has a cache memory. The controller receives commands and data from, and returns at least data to, typically, a plurality of host computers.
A method for measuring system performance in a mass storage system, the storage system having a plurality of disk drive storage elements controlled by a disk drive controller, said controller receiving commands and data from and returning at least data to a plurality of host computers, features enabling a graphical user interface for generating an input parameter containing sequence input of commands for operating the system for measuring system performance, generating from the input parameter sequence a test sequence input identifying commands to be sent to, the storage system, executing at at least one host computer a test request identified by the test sequence input, by sending commands to the mass storage system, accumulating, at at least the executing host computer, data regarding performance of the mass storage system, in response to the requests sent by the host computer, and processing the accumulated data regarding the performance of the mass storage system in response to the one host-generated commands. In one aspect, the processing features validating and correcting, as required, the accumulated data.
The invention further features generating, at the graphical user interface, configuration files, workbench files, and/or benchmark files. In another aspect, the invention features selecting, using the graphical user interface, from various test types for the input sequence of commands, in a point and click fashion. The test types can include defining a system configuration, test periods, and a sequence of test repeat parameters.
Advantageously, therefore, the system easily, and in large part automatically, develops, measures, generates and analyzes statistics describing the dynamic performance of the mass storage system, from the host computers to the disk drive elements, wherein individual operations as well as sequences of operations can easily be set up, with selected initial conditions, and accurately and automatically tested according to desired criteria using graphical user interfaces. The method and apparatus of the invention further advantageously enable the user to easily configure, set, and determine, using a graphical user interface, read/write sequencing and a relative mix of read and write commands, as well as enabling the dynamic performance of the system to be repeatedly enabled and tested for consistency, accuracy, and effectiveness.
Other features and advantages of the invention will be apparent from the following description, taken together with the drawings, in which:
To determine the limits of performance in the system, the hosts can, according to the invention, be operated to exercise and test the memory controller and the disk drive elements. Thus potential problems which can create a bottleneck on those communication lines connected from the disk drive controller to either the disk drive elements or the hosts can be identified, as can cache memory loading issues in the drive controller.
Referring now to
The driver program effects time synchronization of the host computers and causes the next lower level controller programs (the programs which control the commands and data sent to and received from the disk controller), to operate in time synchronization (step 64). In order to achieve both time synchronization of the host computers, and accurate and timely operation of the mass storage system, the driver program may first cause each of the client host computers to synchronize its clock with the master host computer. Further, in response to a communication from the master computer, all host computers begin to issue commands to the controller of the mass storage system, based upon the arguments or parameters previously stored in their memories.
After the necessary test-defining parameters are stored in the memory of each respective host computer, the test is ready to proceed. The next lower level controller program in each of the host computers, designated the scripting program, causes each of the host computers to command the controller in accordance with the command information provided to it by the master host computer driver program (step 66). (Note that the master host computer itself also “receives” such information from the driver program.)
Depending upon the particular disk controller system, one of two possible methods of operation can proceed. If the controller is a “simple” controller, each host computer will itself measure and collect statistics (the raw data) identifying, for example, the response time for each command which it sends to the controller. These response times are collected in a manner which allows the host computer to identify to which command the response time corresponds. Alternatively, for a controller such as the EMC Symmetrix® controller, the controller itself can provide or supplement the raw data for each of the commands which it receives from the hosts. Under this latter circumstance, the controller will return not only the data requested by the command, but in addition, in response to special host requests, the statistics describing the response times for the commands which are being received. That information is provided to the particular host which requested the operation.
Each host computer, then, analyzes the response time data which it either has received or generated itself (step 68). In the illustrated embodiment of the invention, this raw data, which can amount to several gigabytes of information, is analyzed preferably, at each host computer. The analysis is basically a data reduction analysis whereby each host computer, rather than maintaining the full raw data set, operates to reduce the received response times to a smaller set of data. In a first particular embodiment, the data is placed into “buckets”, each bucket representing, for example, 0.25 seconds. (In other embodiments, differently sized buckets can be employed). The buckets, however, collectively represent a non-overlapping, continuous sequence of time durations.
In another embodiment of the invention, the response times can be accumulated for a period of time, so that the data returned to the master host computer will represent the cumulative response times for all commands issued during each of a plurality of non-overlapping contiguous larger durations of time, for example 5 or 10 seconds. That is, for each of the contiguous time periods, the response times for each command initiated in the period will be accumulated. This is particularly useful where the tests can run for several hours and in which tens of gigabytes of data will be produced for each of the host computers.
No matter what method is used to collect and/or reduce the data, the master host computer collects the resulting data from each other host computer (step 70). The master host computer, at the driver program level, then can further analyze the reduced data, as described below, to obtain and present statistics to the user (step 72). These additional statistics can provide further insight and understanding into the performance and operation of portions of or the entire of computer/memory system.
Referring now to the operation of the computer system in more detail, and referring to
Referring now to
Once the arguments have been stored in the data files on that host computer, which is designated as the master computer, the main program invokes the driver program, running on the master computer, to set up all of the hosts. This is indicated at step 106. The driver program, in response to the arguments provided, will initialize the cache memory, if necessary, and will initialize as well, all of the host computers, including the master host computer. The driver program, at 108, causes all of the host computers to be time synchronized. (If time synchronized, it may perform this function by sending to each of the client host computers over channel 50, the clock time of the master host computer.) Each client host computer now runs the scripting program in response to initialization by the driver program and communications from that program over communications channel 50, linking all of the host computers. (Thus all of the client host computer clocks may then be set in some embodiments of the invention, with the result that all of the host computers participating in the test are operating with time synchronized clocks.)
The driver program then transfers to each of the client host computers the necessary configuration and parameter files with which the scripting program, at the client host computers and at the master host computer, will operate to test the mass storage system. This is indicated at step 110.
Next at step 112, the driver program initiates testing of the mass storage system by communicating to each host computer, directly from the master host computer and over the interconnecting communications channel 50. As a result, each host computer begins sending commands and data to and receives at least data from the mass storage system at the same time. At this point, it is the configuration and parameter input to the master host computer, as delivered to the client computers, which controls the actions of each of the client host computers. Thus, the provided information and arguments can cause, for example, only a subset of the host computers to communicate and issue commands to the mass storage system, and/or only a specific set of logical units at the mass storage level may be exercised in a specific configuration dictated by the arguments input at step 100.
The scripting program, when a test is complete, as tested at 114, then reduces the data which it has collected (step 116). As noted above, the data reduction process, if one is used, can use either a bucket compression or an accumulation compression approach. (The approach can be dynamically changed during the test sequence by, for example, a user command to the master host computer.) Alternatively, the initial raw data may be maintained in its original form. The raw data can include, for example, the response times to read and/or write operations, which have been commanded in accordance with the input parameters. (In accordance with the invention, when there is a mix of read and write commands, the system first issues a block of one set of the commands and then a block of the other set of commands. For example, when there are to be 40% read commands, the system can issue three write commands, followed by two read commands, followed by three write commands, etc. In this manner, the associated statistical data which is collected can be directly correlated to a particular read or write command.)
Once the data is in its final (and most likely, reduced) form at the client host computers, it is transferred over channel 50 to the master host computer. This is indicated at step 118. Thereafter the master host computer can effect a more comprehensive data analysis, at step 120, to determine the performance of the mass storage system, for example, the number of I/O's, and in particular, the number of writes, as a function of time. The driver program then determines whether another test is to be performed, at step 104, and if not, the driver checks to determine whether further analysis of the data is to be performed (at step 122). If further analysis data is to be collected using a new configuration, the new configuration is generated at step 124, and the process begins again starting with step 110. If no further analysis data is needed, the system returns to the beginning of the process as indicated by loop 126. In this manner, the three nested loops of this preferred particular embodiment, loops represented by the main program, the driver program, and the scripting program, can provide effective and dynamic testing of the storage system to determine its performance under a large variety of situations.
The data reduction workbench system of the invention begins with a “project tab” which creates or opens a directory structure to allow for management of the project data. This function organizes the data reduction workbench and allows creation and management of new, or the opening of existing, projects and test phases, and further insures consistent naming of all of the components within a project. It also logs all operations performed at both a project and a test phase level. At the test phase level, the system creates or opens the folders for a particular test phase which will hold all of the test data and informational data regarding that particular test. Each test is identified, at 290, and the organization process to effect data transfer begins. Next, in the data transfer phase, at 300, a file transfer protocol is incorporated for transferring the data from the test environment (either the host computers or the disk controller system, such as the EMC Symmetrix® controller system). This mechanism provides the directory structure of all test data and information data for the particular test (benchmark) which has been run.
Once the data has been transferred and is ready to be processed, the system processes the raw data, at 310, by reading the test files and creating summaries and error files for the test. The processing of raw data further creates statistical analysis program files, for example, for further processing and a cache ratio report to ensure that the actual cache ratio is within an acceptable margin of the targeted cache ratio for the characterization workload test. The data is also checked for errors at this stage and those errors are reported, and in some instances, corrected. The coalesced data is then formatted for input to the post processing stage at 320. The coalesced data results from processing the raw data, by automatically processing characterization and database simulation benchmark data, as well as controller generated internal data, and coalescing, according to the test, test iteration, and test configuration, the data collected by each host operating in the test. The coalesced data is thus provided in a single data file for the following post processing.
The post processing phase 320 provides an interface to the statistical analysis data processing program, and to other analysis tools such as the Write Analysis Tool described in U.S. application Ser. No. 09/542,463 entitled BENCHMARK TOOL WRITE ANAYSIS FOR A MASS STORAGE SYSTEM, and filed Apr. 4, 2000, and incorporated herein in its entirety by reference, and allows for plot generation, data summarization, and the importing of the test and information data for a particular test, using a graphical user interface. As a result, formal plot presentations can be automatically loaded into an electronic document, for presentation in a standardized format.
Once post processing has been completed, the system provides interfaces to other presentation formats to generate spreadsheets from the statistical analysis program summary files created during post processing to enable charts and tables to be created in, for example, a standardized word processing report. This enables the development of performance reports. This is indicated at 330. In addition to the reports, this system also automatically loads the benchmark statistical analysis summary data into a database (at 340). With the data thus available, a trend analysis can be created and displayed according to the test types and system configurations. This is indicated at 350. This enables the comparison of performance behaviors across multiple projects (that is, as a function of time) and test phases (for example as a result of different operating software for the same configuration of equipment) in order to determine visually, at first impression, trends which may affect the analysis of system performance.
Once the data is collected, according to the invention, the system organizes, validates, checks, and presents the data as noted below. In the raw data post processing and correction stage, the validity checks can include checking individual data points for errors and applying predetermined decision criteria to either correct, ignore, or delete erroneous data. In addition, an overall validity comparison of the actual (and corrected) results with expected results based on the I/O configuration profile can be effected. The user of pre-established criteria to decide whether to correct, ignore or use the erroneous data is implemented in accordance with one embodiment of the invention by flagging any inconsistencies based on expected results of a particular input/output type and storing such flagged data in a system data files. Once the corrections have been effected, the revised data can be put into any of a number of industry standard templates.
Accordingly, during the post processing phase of operation, in a preferred embodiment of the invention, referring to
The various data files are summarized in FIG. 11. These data files are typically statistical analysis program related files and can be used to create database simulation objects as well as being placed into the correct report format as noted above.
The trend analysis referred to above, taking place as indicated at 350, provides a series of graphs, such as those illustrated in
Referring more particularly to
If the chart selection was “on-line”, the analysis generates the chart on the display screen. The chart can also, as an alternative, for example, be placed in a word processing document by selecting “test type” or “view all”, for example, at the chart selection 1204.
The user selects the test phases which are to be graphed together. By default, the project and one test phase will be selected, however, all test phases for a project can be selected by clicking, for example, on the project. Alternatively, specific desired test phases can be selected (or deselected) as needed. As noted above, the graphs can be generated and viewed online, or printed, and can be shown according to test type or types as required. The system also, by pressing the “scale” selection 1206, will calculate the minimum and maximum values for the data and use those values to scale the graph. A percent scale is also available where the largest value of the chart is considered at 100% and all of the values are displayed accordingly.
The various specific information is shown in the characterization frame 1208 and is selected by the user and viewed as a chart online or report.
During the input process, the user selects, using the graphical user interface, the parameters which enable the graphs of
By plotting this information in a graphical format, the user is enabled to spot trends in the data as a result of changes over time, or other parameters. This data is also available for viewing on screen.
Graphical User Interface
The system provides for a graphical user interface which enables operational parameters of the system to be created quickly, with repeatability, reliability, and use by a much broader audience. The graphic user interface is essentially a front-end device for the invention which automatically operates to generate and/or work on three types of files: configuration rule data, workload data, and benchmark data. It allows the selection of various test types based upon the user inputs which are provided in a “point and click” manner. The graphic user interface of the invention is substantially more reliable for the user and enables the user to quickly and easily define the system tests which are to be performed.
The typical graphic user interface is presented, in accordance with the invention, in
The user is thus enabled to form this presentation using the graphical user interface.
Additions, subtractions, and other modifications of the illustrated embodiment of the invention will be apparent to those practicing in this field and are within the scope of the following claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5678046 *||May 5, 1995||Oct 14, 1997||The Chase Manhattan Bank, N.A.||Method and apparatus for distributing files on a file storage device|
|US5940855 *||Apr 1, 1998||Aug 17, 1999||Sun Microsystems, Inc.||System and method for determining relative cache performance in a computer system|
|US5953689 *||Mar 12, 1998||Sep 14, 1999||Emc Corporation||Benchmark tool for a mass storage system|
|US6098128 *||Sep 17, 1996||Aug 1, 2000||Cyberstorage Systems Corporation||Universal storage management system|
|US6260120 *||Jun 29, 1998||Jul 10, 2001||Emc Corporation||Storage mapping and partitioning among multiple host processors in the presence of login state changes and host controller replacement|
|US6295575 *||Jun 29, 1998||Sep 25, 2001||Emc Corporation||Configuring vectors of logical storage units for data storage partitioning and sharing|
|US6301605 *||Sep 29, 2000||Oct 9, 2001||Adaptec, Inc.||File array storage architecture having file system distributed across a data processing platform|
|US6330554 *||Jun 3, 1999||Dec 11, 2001||Microsoft Corporation||Methods and apparatus using task models for targeting marketing information to computer users based on a task being performed|
|US6487562 *||Dec 20, 1999||Nov 26, 2002||Emc Corporation||Dynamically modifying system parameters in data storage system|
|US20020002606 *||Mar 6, 1998||Jan 3, 2002||David H. Jaffe||Method and system for managing storage devices over a network|
|US20020026549 *||Jul 31, 2001||Feb 28, 2002||Powers Tracy D.||Remote reconfiguration system|
|JPH11288404A *||Title not available|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7093011||Apr 28, 2003||Aug 15, 2006||Hitachi, Ltd.||Storage network system, managing apparatus, managing method and program|
|US7133795||Dec 23, 2004||Nov 7, 2006||Emc Corporation||Techniques for testing an electronic system using in-line signal distortion|
|US7171338 *||Aug 18, 2000||Jan 30, 2007||Emc Corporation||Output performance trends of a mass storage system|
|US7293116||Mar 24, 2005||Nov 6, 2007||Xiv Ltd.||Graphic user interface for a storage system|
|US7412504||Aug 4, 2005||Aug 12, 2008||Hitachi, Ltd.||Storage network system, managing apparatus managing method and program|
|US7412506||Apr 3, 2006||Aug 12, 2008||Hitachi, Ltd.||Storage network system, managing apparatus managing method and program|
|US7477949||May 8, 2006||Jan 13, 2009||Xiv Ltd.||Graphic user interface for a storage system|
|US7610369||Oct 27, 2009||Hitachi, Ltd.||Storage network system, managing apparatus managing method and program|
|US7730358 *||Oct 24, 2007||Jun 1, 2010||Inventec Corporation||Stress testing method of file system|
|US7987256||Sep 4, 2009||Jul 26, 2011||Hitachi, Ltd.||Storage network system, managing apparatus, managing method and program|
|US8055686||Nov 8, 2011||Hitachi, Ltd.||Method and program of collecting performance data for storage network|
|US8082338||Dec 20, 2011||Hitachi, Ltd.||Storage network system, managing apparatus, managing method and program|
|US8095332 *||Jan 10, 2012||Getac Technology Corporation||Method for pre-heating hard disk drive of computer device|
|US8171126||May 1, 2012||Hitachi, Ltd.||Storage network system, managing apparatus, managing method and program|
|US8230057||Jul 24, 2012||Hitachi, Ltd.||Storage network system, managing apparatus, managing method and program|
|US8239640 *||Aug 7, 2012||Dataram, Inc.||System for controlling performance aspects of a data storage and access routine|
|US8549050||Sep 22, 2011||Oct 1, 2013||Hitachi, Ltd.||Method and system for collecting performance data for storage network|
|US8856264 *||Oct 28, 2011||Oct 7, 2014||Hitachi, Ltd.||Computer system and management system therefor|
|US9336155 *||Aug 31, 2015||May 10, 2016||Facebook, Inc.||Statistical cache promotion|
|US20040024870 *||Apr 28, 2003||Feb 5, 2004||Hitachi, Ltd.||Storage network system, managing apparatus, managing method and program|
|US20060015605 *||Aug 4, 2005||Jan 19, 2006||Toshiaki Hirata||Storage network system, managing apparatus managing method and program|
|US20060171334 *||Apr 3, 2006||Aug 3, 2006||Hitachi, Ltd.||Storage network system, managing apparatus managing method and program|
|US20060214936 *||Mar 24, 2005||Sep 28, 2006||Xiv Ltd.||Graphic user interface for a storage system|
|US20060267984 *||May 8, 2006||Nov 30, 2006||Ofir Zohar||Graphic user interface for a storage system|
|US20080249885 *||Apr 3, 2007||Oct 9, 2008||Yahoo! Inc.||Systems and methods for providing an internet asset marketplace|
|US20090113249 *||Oct 24, 2007||Apr 30, 2009||Inventec Corporation||Stress testing method of file system|
|US20100061011 *||Mar 11, 2010||Yi-Chung Hsien||Method for pre-heating hard disk drive of computer device|
|US20100082806 *||Apr 1, 2010||Hitachi, Ltd.||Storage network system, managing apparatus, managing method and program|
|US20100095073 *||Oct 9, 2008||Apr 15, 2010||Jason Caulkins||System for Controlling Performance Aspects of a Data Storage and Access Routine|
|US20110238831 *||Sep 29, 2011||Hitachi, Ltd.||Storage network system, managing apparatus, managing method and program|
|US20130110966 *||Oct 28, 2011||May 2, 2013||Hitachi, Ltd.||Computer system and management system therefor|
|US20150302034 *||Apr 17, 2014||Oct 22, 2015||Netapp, Inc.||Correlating database and storage performance views|
|USD756397 *||Apr 4, 2014||May 17, 2016||Adp, Llc||Display screen or portion thereof with a graphical user interface|
|USD757077 *||Aug 27, 2014||May 24, 2016||Robert Bosch Gmbh||Display screen with a graphical user interface|
|USD757078 *||Aug 27, 2014||May 24, 2016||Robert Bosch Gmbh||Display screen with a graphical user interface|
|USD757776 *||Aug 27, 2014||May 31, 2016||Robert Bosch Gmbh||Display screen with a graphical user interface|
|USD759075 *||Apr 11, 2014||Jun 14, 2016||Nutonian, Inc.||Display screen with graphical user interface|
|USD759076 *||Apr 18, 2014||Jun 14, 2016||Nutonian, Inc.||Display screen with graphical user interface|
|U.S. Classification||702/186, 702/182, 702/122, 714/E11.182, 714/E11.206, 714/E11.193, 702/123, 710/6|
|International Classification||G06F13/00, G06F11/32, G06F11/34|
|Cooperative Classification||G06F11/3414, G06F11/3485, G06F11/321|
|European Classification||G06F11/32D, G06F11/34T8, G06F11/34C2|
|Aug 18, 2000||AS||Assignment|
Owner name: EMC CORPORATION, MASSACHUSETTS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:LALLY, MAUREEN A.;GOGUEN, KENNETH R.;LAUTERBACH, PETER C.;REEL/FRAME:011053/0460
Effective date: 20000816
|May 24, 2005||CC||Certificate of correction|
|Jul 18, 2008||FPAY||Fee payment|
Year of fee payment: 4
|Jul 18, 2012||FPAY||Fee payment|
Year of fee payment: 8